Regional campaing’s results (PhD Chapter 1)

This series of files compile all analyses done during Chapter 1 for the regional campaign (2016):

All analyses have been done with PRIMER-e 6 and R 3.6.0.

Click on the table of contents in the left margin to assess a specific analysis.
Click on a figure to zoom it

To assess maps and figures, click here.
To go back to the summary page, click here.

Caracteristics of each campaign

2014 2016 2017
Sampling date August-September June to August July
Criteria for perturbation Potentially impacted if close to the city or industries, References outside the bay Human-impacted if in a region with a highly populated area, industries and maritime activities, Reference if none of these criteria Human-impacted if in a region with a highly populated area, industries and maritime activities, Reference if none of these criteria
Regions considered BSI BSI, CPC, BDA, MR BSI, MR
Number of sampled stations 40 (20 HI, 20 R) 78 (26 BSI, 19 CPC, 18 BDA, 15 MR) 126 (111 BSI, 15 MR)
Parameters sampled Organic matter yes yes yes
Photosynthetic pigments no yes yes
Sediment grain-size yes yes yes
Heavy-metals yes yes (for a limited number of stations) no (interpolated based on 2014 and 2016 values)
Benthic communities Compartment targeted Macro-infauna Macro-infauna Macro-infauna
Sieved used 500 µm 1 mm 500 µm and 1 mm
Conservation technique Formaldehyle Formaldehyle Formaldehyle
Others N.A. N.A. N.A.

We used data from subtidal ecosystems (see metadata files for more information). Only stations that have been sampled both for abiotic parameters and benthic species were included.

Selected variables for the analyses:

Abundances of Mesodesma arctatum (Marc) and Cistenides granulata (Cgra) were also considered (see IndVal and SIMPER results).

As data is missing for metal concentrations outside BSI, two Designs have been used:

Statistics for each variable considered:

BSI, CPC, BDA, PR (habitat parameters)
  Min Max Median Mean SD SE 95% CI
depth 1.000 65.900 16.100 21.131 17.504 1.982 3.884
om 0.168 3.863 0.387 0.689 0.730 0.083 0.162
gravel 0.000 0.809 0.000 0.049 0.122 0.014 0.027
sand 0.000 1.001 0.873 0.707 0.348 0.039 0.077
silt 0.000 0.942 0.031 0.200 0.285 0.032 0.063
clay 0.000 0.497 0.006 0.043 0.098 0.011 0.022
S 0.000 13.000 4.000 4.795 2.969 0.336 0.659
N 0.000 142.000 12.000 19.846 25.867 2.929 5.740
H 0.000 2.307 1.069 1.080 0.610 0.069 0.135
J 0.000 1.000 0.838 0.724 0.292 0.033 0.065
BSI (heavy metals)
  Min Max Median Mean SD SE 95% CI
arsenic 0.800 21.300 2.750 4.188 4.582 0.899 1.761
cadmium 0.080 0.270 0.145 0.152 0.047 0.009 0.018
chromium 17.000 111.000 42.500 49.385 23.565 4.621 9.058
copper 2.400 28.800 6.450 9.373 7.593 1.489 2.919
iron 21938.100 85408.600 42304.650 45356.633 16635.601 3262.510 6394.402
manganese 318.400 1657.100 711.450 776.662 348.066 68.261 133.790
mercury 0.007 0.091 0.019 0.023 0.017 0.003 0.006
lead 1.900 12.200 4.750 5.381 2.902 0.569 1.115
zinc 26.900 141.400 46.800 58.262 29.368 5.760 11.289

1. Permutational Analyses of Covariance

Results of univariate PermANCOVAs on parameters and multivariate PermANCOVA on the whole benthic community with depth as covariate are presented in the table below.

Variable Condition Region(Co) Depth Significative groups of similar regions (p > 0.05)
om S S (CPC BDA MR)
gravel All regions in the same group
sand S All regions in the same group
silt S S (BSI CPC BDA), (BDA MR)
clay (BSI BDA MR), (CPC MR)
S (1 mm) S (BSI CPC MR), (CPC BDA MR)
N (1 mm) All regions in the same group
H (1 mm) S S (CPC BDA MR), (BSI MR)
J (1 mm) (BSI CPC MR), (CPC BDA MR)
ALL SPECIES (1 mm) S S

2. IndVal and SIMPER

These analyses allowed to select species as dependant variables for the regressions. We used results from PRIMER to justify further their choice.

##                       cluster indicator_value probability
## cistenides_granulata        1          0.3080       0.018
## ennucula_tenuis             1          0.2222       0.006
## macoma_calcarea             1          0.2222       0.004
## eudorellopsis_integra       1          0.1556       0.038
## mesodesma_arctatum          2          0.2535       0.003
## harmothoe_imbricata         2          0.2008       0.005
## glycera_alba                2          0.1212       0.029
## psammonyx_nobilis           2          0.1212       0.036
## 
## Sum of probabilities                 =  51.022 
## 
## Sum of Indicator Values              =  6.08 
## 
## Sum of Significant Indicator Values  =  1.6 
## 
## Number of Significant Indicators     =  8 
## 
## Significant Indicator Distribution
## 
## 1 2 
## 4 4
SIMPER results (average dissimilarity: 94.67 )
  average sd ratio ava avb cumsum
echinarachnius_parma 0.131 0.205 0.638 4.18 1.33 0.138
mesodesma_arctatum 0.0987 0.205 0.482 3.52 0.267 0.242
cistenides_granulata 0.0671 0.128 0.525 0.485 1.62 0.313
strongylocentrotus_sp 0.0438 0.108 0.404 0.667 0.867 0.36
nephtys_caeca 0.034 0.0584 0.581 0.636 0.667 0.395
protomedeia_grandimana 0.0329 0.0819 0.402 0.636 0.933 0.43
scoloplos_armiger 0.0317 0.0798 0.397 1.06 1.11 0.464
amphipholis_squamata 0.0306 0.119 0.257 0.0606 2.96 0.496
limecola_balthica 0.0266 0.0543 0.49 0.606 0.489 0.524
macoma_calcarea 0.0241 0.0588 0.41 0 0.733 0.549
thyasira_sp 0.0233 0.0613 0.38 0.0303 0.933 0.574
ennucula_tenuis 0.0225 0.0545 0.413 0 1.02 0.598
psammonyx_nobilis 0.0203 0.0744 0.273 0.515 0 0.619
harmothoe_imbricata 0.0193 0.052 0.372 0.394 0.0222 0.64
pygospio_elegans 0.0193 0.0996 0.193 2.76 0.0222 0.66
pontoporeia_femorata 0.0151 0.0727 0.208 0 0.733 0.676
glycera_alba 0.0142 0.0566 0.25 0.515 0 0.691
ameritella_agilis 0.0141 0.0687 0.206 0 0.444 0.706
astarte_undata 0.014 0.0482 0.29 0.364 0.0222 0.721
ciliatocardium_ciliatum 0.0136 0.0497 0.273 0.182 0.222 0.735
astarte_subaequilatera 0.013 0.0485 0.268 0.364 0 0.749
bipalponephtys_neotena 0.0122 0.0715 0.17 0 0.556 0.762
mya_arenaria 0.0114 0.0264 0.433 0.0909 0.311 0.774
goniada_maculata 0.0109 0.0382 0.286 0.0303 0.333 0.785
ampharete_oculata 0.0104 0.0649 0.16 0.242 0 0.796
nucula_proxima 0.0096 0.0398 0.241 0 0.244 0.806
glycera_dibranchiata 0.00933 0.0351 0.266 0.0303 0.133 0.816
diastylis_sculpta 0.00903 0.0473 0.191 0.121 0.0444 0.826
testudinalia_testudinalis 0.00864 0.0382 0.226 0.212 0.0444 0.835
ampharetidae_spp 0.0069 0.0212 0.326 0.121 0.111 0.842
eudorellopsis_integra 0.00677 0.0217 0.312 0 0.333 0.849
maldanidae_spp 0.00645 0.0231 0.279 0.212 0.0444 0.856
ampeliscidae_spp 0.00641 0.0199 0.322 0.0909 0.0889 0.863
yoldia_myalis 0.00622 0.0215 0.289 0.0909 0.0667 0.87
nephtys_ciliata 0.00618 0.0252 0.245 0 0.2 0.876
ophelia_limacina 0.00592 0.0203 0.292 0.0606 0.0889 0.882
phyllodoce_mucosa 0.00572 0.0232 0.247 0 0.333 0.888
polynoidae_spp 0.00571 0.0156 0.366 0.0303 0.222 0.894

3. Univariate regressions

Independant variables are habitat parameters and heavy metal concentrations, dependant variables are diversity indices. We used linear models for the all regressions on diversity indices.

3.1. Identification of outliers

To identify stations that are not consistent with the others, we used the multivariate Cook’s Distance (CD) on the uncorrelated variables. A significative threshold of 4 times the mean of CD has been established.

Design 1

Based on Cook’s Distance, we identified stations 80, 96, 107 and 126 as general outliers. They have been deleted for the following analyses of Design 1.

Design 2

Based on Cook’s Distance, we identified stations 108 and 110 as general outliers. They have been deleted for the following analyses of Design 2.

3.2. Correlations between parameters

Correlations have been calculated with Spearman’s rank coefficient.

Design 1

Correlation coefficients between habitat parameters (Design 1)
  om gravel sand silt clay
om 1 -0.045 -0.823 0.737 0.731
gravel -0.045 1 -0.186 -0.347 -0.296
sand -0.823 -0.186 1 -0.793 -0.799
silt 0.737 -0.347 -0.793 1 0.972
clay 0.731 -0.296 -0.799 0.972 1

According to these results, the following variables are highly correlated (\(|\rho|\) > 0.80) so they have been considered together in the regressions of Design 1:

  • silt and clay (clay deleted)

We decided to keep sand, even if it is correlated with om, to stay consistant with the local campaign.

Design 2

Correlation coefficients between heavy metals concentrations (Design 2)
  arsenic cadmium chromium copper iron manganese mercury lead zinc
arsenic 1 0.476 0.729 0.835 0.772 0.389 0.626 0.819 0.888
cadmium 0.476 1 0.731 0.421 0.717 0.825 0.108 0.675 0.66
chromium 0.729 0.731 1 0.665 0.799 0.753 0.439 0.84 0.852
copper 0.835 0.421 0.665 1 0.56 0.296 0.518 0.805 0.873
iron 0.772 0.717 0.799 0.56 1 0.728 0.395 0.71 0.807
manganese 0.389 0.825 0.753 0.296 0.728 1 0.074 0.546 0.565
mercury 0.626 0.108 0.439 0.518 0.395 0.074 1 0.61 0.505
lead 0.819 0.675 0.84 0.805 0.71 0.546 0.61 1 0.881
zinc 0.888 0.66 0.852 0.873 0.807 0.565 0.505 0.881 1

According to these results, the following variables are highly correlated (\(|\rho|\) > 0.80) so they have been considered together in the regressions of Design 2:

  • cadmium and manganese (manganese deleted)
  • copper, lead and zinc (copper and zinc deleted)

We decided to keep arsenic, even though it is correlated with the copper/lead/zinc group, to stay consistant with the local campaign.

3.3. Simple regressions

These analyses have been do to explore the relationships between variables. As it is a huge number of results to interpret, only multiple regressions will be included in the article (see below).

Design 1

Adjusted R-squared of simple regressions for Design 1
  om gravel sand silt
S 0.09191 0.01477 0.05878 0.1146
N 0.01404 0.002513 0.03299 0.03504
H 0.0883 -0.01004 0.05319 0.0994
J -0.0001471 0.000586 0.01035 0.009213
p-values of simple regressions for Design 1
  om gravel sand silt
S 0.004997 0.1522 0.02111 0.001855
N 0.1576 0.2802 0.06578 0.06001
H 0.005847 0.602 0.02694 0.003605
J 0.3233 0.3106 0.1884 0.1992

Design 2

Adjusted R-squared of simple regressions for Design 2
  arsenic cadmium chromium iron mercury lead
S -0.02448 -0.04543 -0.03355 -0.04355 -0.04506 0.0487
N -0.007642 -0.0415 -0.03362 -0.04259 -0.04478 0.02349
H -0.02186 -0.03955 -0.008538 -0.02448 -0.0236 0.08028
J -0.04475 -0.02601 -0.02849 -0.02868 -0.02868 -0.04439
p-values of simple regressions for Design 2
  arsenic cadmium chromium iron mercury lead
S 0.5091 0.9838 0.6196 0.8432 0.9284 0.1542
N 0.3734 0.7753 0.6207 0.8081 0.9061 0.2258
H 0.4835 0.7271 0.3792 0.5091 0.5003 0.09687
J 0.9043 0.5252 0.5531 0.5554 0.5553 0.8824

Furthermore, depth has been shown important for several parameters in the ANCOVAs, so here are the corresponding scatterplots.

3.4. Multiple regressions

This section presents analyses done (i) to determine which model (Design 1, Design 2) decribes the best the parameters and (ii) which variables are the most important to explain the parameters.

3.4.1. Best model selection

This step was not used here as both models were needed.

3.4.2. Significative variables selection

We identified which variables were selected after an AIC procedure to predict the best the parameters. Results of the variable selection, according to AIC, are shown on the tables below:

  • for the model of Design 1
Variable (or combination) S N H J
om
gravel - + +
sand + - + +
silt/clay + - + +
Adjusted \(R^{2}\) 0.13 0.06 0.15 0.04
  • for the model of Design 2
Variable (or combination) S N H J
arsenic
cadmium/manganese
chromium - - -
iron
mercury -
lead/copper/zinc + + +
Adjusted \(R^{2}\) 0.27 0.12 0.18 0

Details of the regressions, with diagnostics and cross-validation, are summarized below.

Design 1

Species richness
## FULL MODEL
## Adjusted R2 is: 0.11
Fitting linear model: S ~ om + gravel + sand + silt
  Estimate Std. Error t value Pr(>|t|)
(Intercept) -1.002 4.832 -0.2074 0.8363
om 0.3518 0.7513 0.4683 0.6411
gravel 2.191 5.784 0.3788 0.706
sand 5.129 4.869 1.053 0.2959
silt 8.536 6.459 1.321 0.1907
Variance Inflation Factors
  om gravel sand silt
VIF 1.83 2.15 5.34 5.96 
## REDUCED MODEL
## Adjusted R2 is: 0.13
Fitting linear model: S ~ sand + silt
  Estimate Std. Error t value Pr(>|t|)
(Intercept) 0.8442 2.242 0.3765 0.7076
sand 3.406 2.349 1.45 0.1514
silt 7.234 2.793 2.59 0.01163 *
Variance Inflation Factors
  sand silt
VIF 2.6 2.6 
## RMSE for the full model: 2.674811 
## RMSE for the reduced model: 2.605987

Total abundance
## FULL MODEL
## Adjusted R2 is: 0.05
Fitting linear model: N ~ om + gravel + sand + silt
  Estimate Std. Error t value Pr(>|t|)
(Intercept) 108.4 47.04 2.304 0.02423 *
om 3.953 7.314 0.5405 0.5906
gravel -116.2 56.31 -2.063 0.04287 *
sand -92.11 47.4 -1.943 0.05606
silt -97.86 62.88 -1.556 0.1242
Variance Inflation Factors
  om gravel sand silt
VIF 1.83 2.15 5.34 5.96 
## REDUCED MODEL
## Adjusted R2 is: 0.06
Fitting linear model: N ~ gravel + sand + silt
  Estimate Std. Error t value Pr(>|t|)
(Intercept) 102.3 45.43 2.251 0.02751 *
gravel -105.4 52.41 -2.011 0.04814 *
sand -85.09 45.35 -1.876 0.0648
silt -80.89 54.21 -1.492 0.1402
Variance Inflation Factors
  gravel sand silt
VIF 2.01 5.13 5.16 
## RMSE for the full model: 28.13056 
## RMSE for the reduced model: 27.4413

Shannon index
## FULL MODEL
## Adjusted R2 is: 0.13
Fitting linear model: H ~ om + gravel + sand + silt
  Estimate Std. Error t value Pr(>|t|)
(Intercept) -1.323 0.977 -1.355 0.18
om 0.01376 0.1519 0.09059 0.9281
gravel 2.252 1.17 1.926 0.05827
sand 2.273 0.9846 2.308 0.02398 *
silt 3.306 1.306 2.531 0.01365 *
Variance Inflation Factors
  om gravel sand silt
VIF 1.83 2.15 5.34 5.96 
## REDUCED MODEL
## Adjusted R2 is: 0.15
Fitting linear model: H ~ gravel + sand + silt
  Estimate Std. Error t value Pr(>|t|)
(Intercept) -1.345 0.9417 -1.428 0.1578
gravel 2.29 1.086 2.108 0.03864 *
sand 2.297 0.9401 2.444 0.01706 *
silt 3.365 1.124 2.995 0.003795 * *
Variance Inflation Factors
  gravel sand silt
VIF 2.01 5.13 5.16 
## RMSE for the full model: 0.5407726 
## RMSE for the reduced model: 0.5336406

Piélou’s evenness
## FULL MODEL
## Adjusted R2 is: 0.03
Fitting linear model: J ~ om + gravel + sand + silt
  Estimate Std. Error t value Pr(>|t|)
(Intercept) -0.1486 0.4941 -0.3007 0.7646
om -0.06691 0.07683 -0.8709 0.3868
gravel 1.28 0.5915 2.165 0.03389 *
sand 0.8545 0.4979 1.716 0.09064
silt 1.322 0.6606 2.001 0.04933 *
Variance Inflation Factors
  om gravel sand silt
VIF 1.83 2.15 5.34 5.96 
## REDUCED MODEL
## Adjusted R2 is: 0.04
Fitting linear model: J ~ gravel + sand + silt
  Estimate Std. Error t value Pr(>|t|)
(Intercept) -0.04525 0.4788 -0.09449 0.925
gravel 1.098 0.5524 1.988 0.05069
sand 0.7355 0.478 1.539 0.1284
silt 1.035 0.5714 1.811 0.07447
Variance Inflation Factors
  gravel sand silt
VIF 2.01 5.13 5.16 
## RMSE for the full model: 0.2730003 
## RMSE for the reduced model: 0.2724726

Design 2

Species richness
## FULL MODEL
## Adjusted R2 is: 0.21
Fitting linear model: S ~ arsenic + cadmium + chromium + iron + mercury + lead
  Estimate Std. Error t value Pr(>|t|)
(Intercept) 8.579 2.427 3.536 0.002541 * *
arsenic -0.1901 0.2538 -0.7492 0.464
cadmium -13.98 22.35 -0.6253 0.5401
chromium -0.1582 0.1052 -1.504 0.1511
iron -3.955e-05 0.0001018 -0.3885 0.7025
mercury -62.83 40.07 -1.568 0.1353
lead 2.234 0.6924 3.227 0.00495 * *
Variance Inflation Factors
  arsenic cadmium chromium iron mercury lead
VIF 2.1 1.78 3.63 2.74 1.21 3.22 
## REDUCED MODEL
## Adjusted R2 is: 0.27
Fitting linear model: S ~ chromium + mercury + lead
  Estimate Std. Error t value Pr(>|t|)
(Intercept) 7.749 1.623 4.774 0.0001157 * * *
chromium -0.2192 0.07724 -2.838 0.01016 *
mercury -48.82 36.49 -1.338 0.1959
lead 1.996 0.6019 3.316 0.003444 * *
Variance Inflation Factors
  chromium mercury lead
VIF 2.77 1.14 2.91 
## RMSE for the full model: 3.363616 
## RMSE for the reduced model: 2.736833

Total abundance
## FULL MODEL
## Adjusted R2 is: 0.02
Fitting linear model: N ~ arsenic + cadmium + chromium + iron + mercury + lead
  Estimate Std. Error t value Pr(>|t|)
(Intercept) 35.33 16.58 2.131 0.04799 *
arsenic 0.3792 1.734 0.2187 0.8295
cadmium 22.2 152.7 0.1453 0.8861
chromium -0.795 0.719 -1.106 0.2843
iron -0.0005012 0.0006957 -0.7205 0.481
mercury -317.9 273.8 -1.161 0.2616
lead 9.857 4.731 2.084 0.05261
Variance Inflation Factors
  arsenic cadmium chromium iron mercury lead
VIF 2.1 1.78 3.63 2.74 1.21 3.22 
## REDUCED MODEL
## Adjusted R2 is: 0.12
Fitting linear model: N ~ chromium + lead
  Estimate Std. Error t value Pr(>|t|)
(Intercept) 25.3 10.34 2.446 0.02332 *
chromium -0.921 0.5068 -1.817 0.08346
lead 8.173 3.76 2.174 0.04133 *
Variance Inflation Factors
  chromium lead
VIF 2.7 2.7 
## RMSE for the full model: 27.13628 
## RMSE for the reduced model: 17.79233

Shannon index
## FULL MODEL
## Adjusted R2 is: 0.07
Fitting linear model: H ~ arsenic + cadmium + chromium + iron + mercury + lead
  Estimate Std. Error t value Pr(>|t|)
(Intercept) 1.403 0.4344 3.229 0.004935 * *
arsenic -0.04832 0.04543 -1.064 0.3023
cadmium -2.595 4.002 -0.6485 0.5253
chromium -0.02205 0.01884 -1.17 0.258
iron 3.069e-06 1.823e-05 0.1684 0.8683
mercury -5.478 7.173 -0.7637 0.4555
lead 0.3167 0.1239 2.555 0.02051 *
Variance Inflation Factors
  arsenic cadmium chromium iron mercury lead
VIF 2.1 1.78 3.63 2.74 1.21 3.22 
## REDUCED MODEL
## Adjusted R2 is: 0.18
Fitting linear model: H ~ chromium + lead
  Estimate Std. Error t value Pr(>|t|)
(Intercept) 1.366 0.2683 5.091 4.833e-05 * * *
chromium -0.0252 0.01315 -1.916 0.06901
lead 0.2401 0.09757 2.461 0.02261 *
Variance Inflation Factors
  chromium lead
VIF 2.7 2.7 
## RMSE for the full model: 0.5236763 
## RMSE for the reduced model: 0.5011726

Piélou’s evenness
## FULL MODEL
## Adjusted R2 is: -0.18
Fitting linear model: J ~ arsenic + cadmium + chromium + iron + mercury + lead
  Estimate Std. Error t value Pr(>|t|)
(Intercept) 0.5817 0.1962 2.964 0.008691 * *
arsenic -0.009212 0.02052 -0.4489 0.6592
cadmium 0.03793 1.808 0.02098 0.9835
chromium 0.006516 0.00851 0.7657 0.4544
iron 3.404e-06 8.234e-06 0.4134 0.6845
mercury 2.582 3.24 0.7968 0.4365
lead -0.05364 0.05599 -0.9579 0.3515
Variance Inflation Factors
  arsenic cadmium chromium iron mercury lead
VIF 2.1 1.78 3.63 2.74 1.21 3.22 
## REDUCED MODEL
## Adjusted R2 is: 0
Fitting linear model: J ~ 1
  Estimate Std. Error t value Pr(>|t|)
(Intercept) 0.7995 0.04169 19.17 1.21e-15 * * *

Quitting from lines 438-442 (C1_analyses_reg2.Rmd) Error in Qr$qr[p1, p1, drop = FALSE] : indice hors limites De plus : There were 30 warnings (use warnings() to see them)

## RMSE for the full model: 0.3017948 
## RMSE for the reduced model: 0.2192119

4. Multivariate regressions

Independant variables are habitat parameters or heavy metal concentrations, dependant variables are species abundances. Outliers and correlated variables have been excluded from the analysis.

This analysis has been done on PRIMER, with a DistLM to identify the variables that explain the most the community variability and with a dbRDA to plot the results.

Design 1

Design 2

Elliot Dreujou

2019-12-11